Bulletin of the World Health Organization

Universal combination antiretroviral regimens to prevent mother-to-child transmission of HIV in rural Zambia: a two-round cross-sectional study

Benjamin H Chi a, Patrick Musonda b, Mwila K Lembalemba c, Namwinga T Chintu d, Matthew G Gartland e, Saziso N Mulenga d, Maximillian Bweupe c, Eleanor Turnbull d, Elizabeth M Stringer a & Jeffrey SA Stringer a

a. University of North Carolina at Chapel Hill School of Medicine, Campus Box 7570, 130 Farm Mason Road, Chapel Hill, NC 27599, United States of America (USA).
b. Department of Medical Statistics, University of East Anglia, Norwich, England.
c. Zambian Ministry of Health, Lusaka, Zambia.
d. Centre for Infectious Disease Research in Zambia, Lusaka, Zambia.
e. Vanderbilt University School of Medicine, Nashville, USA.

Correspondence to Benjamin H Chi (email: bchi@med.unc.edu).

(Submitted: 05 September 2013 – Revised version received: 28 February 2014 – Accepted: 04 March 2014 – Published online: 05 June 2014.)

Bulletin of the World Health Organization 2014;92:582-592. doi: http://dx.doi.org/10.2471/BLT.13.129833

Introduction

In recent years, studies have shown unequivocally that the use of combination antiretroviral regimens, for either treatment or prophylaxis, during the antenatal, intrapartum and breastfeeding periods in women with human immunodeficiency virus (HIV) infections can reduce the rate of transmission to their infants to less than  5%.13 The World Health Organization (WHO) has endorsed this approach to the prevention of mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV), which it terms the Option B strategy. It is relatively simple, is aligned with adult HIV treatment guidelines, is associated with a reduction in comorbid conditions related to acquired immunodeficiency syndrome (AIDS), and can limit HIV transmission to uninfected partners.4,5 In fact, this approach has been extended in many settings into what has been termed Option B+, where antiretroviral therapy (ART) is started during pregnancy for all HIV-infected pregnant women and continued for life.6

Although maternal antiretroviral regimens have been shown to be highly effective in clinical trials, the results are difficult to replicate fully in real world settings,7 largely because of inefficiencies in health systems.8 For example, incomplete uptake of health-care services in a community can result in poor coverage of proven biomedical interventions.9,10 In addition, delays in screening patients for their eligibility for antiretroviral prophylaxis or treatment can result in late drug administration.11 There is also growing evidence in the literature suggesting that, after therapies have been started, treatment adherence and programme retention may be poor among HIV-infected pregnant and breastfeeding women.1214

Research on the implementation of antiretroviral programmes for PMTCT is needed to extend our understanding of the efficacy of these interventions in individuals to their effectiveness in the general population.15,16 This broader understanding is urgently needed by policy-makers and programme managers who are planning to adopt or optimize maternal combination antiretroviral regimens for PMTCT in various settings.6,17,18 The aim of this study was to evaluate, at the population level, a pilot PMTCT programme similar to Option B in rural Zambia (Table 1). Cross-sectional household surveys were performed before and after implementation of the programme to determine whether early childhood survival in the community changed during this time.

Methods

The pilot programme

In April 2009, we implemented a pilot PMTCT programme at four health-care facilities in the rural Kafue district of Zambia using a phased approach. With support from the Zambian Ministry of Health, we offered standard combination antiretroviral regimens to all HIV-infected pregnant women, irrespective of immunological status. A detailed description of the clinical services provided has been presented elsewhere.19 Briefly, consistent with the standard of care in Zambia, all pregnant women who sought antenatal care were offered opt-out HIV counselling and testing; those who tested positive underwent clinical and immunological (i.e. CD4+ T-cell count) screening. These evaluations were required by national treatment guidelines and were used to determine whether the maternal ART should be continued after breastfeeding. In addition, since virological monitoring was not routinely available, these baseline data were also important for assessing treatment responses. Women who met Zambian national guideline criteria for HIV treatment (i.e. a CD4+ T-cell count less than 350 cells/µl or WHO clinical stage 3 or 4) immediately started lifelong antiretroviral therapy.20

In our programme, we also offered three-drug combination regimens to women who did not meet these eligibility criteria, starting at 28 weeks’ gestation and continuing until the cessation of breastfeeding. The regimens were based on the nucleoside reverse transcriptase inhibitors zidovudine and lamivudine, which were combined with either nevirapine or efavirenz. (At the time, Zambian PMTCT guidelines recommended zidovudine monotherapy for HIV-infected pregnant women from 28 weeks’ gestation until delivery and nevirapine peripartum for mother and neonate but no antiretrovirals during breastfeeding.)21 Although we implemented this pilot programme before 2010, when WHO first recommended antiretroviral prophylaxis during breastfeeding, this regimen closely resembles Option B, as described in later WHO guidelines (Table 1).4

Household surveys

To measure changes associated with the PMTCT programme at the population level, we conducted two household surveys in the catchment areas of the four health-care facilities taking part. We used the methods first introduced in the four-country study to measure the 24-month HIV-free survival in children born to HIV-infected mothers.22 We assessed survival using methods similar to those used in Demographic and Health Surveys conducted in many African countries.23 The sampling method was established in the first survey, which was conducted between November 2008 and May 2009, before implementation of the pilot PMTCT programme. Government-demarcated catchment areas were mapped and we randomly selected zones within these communities. These zones were used by each health-care facility’s Neighbourhood Health Committee for community outreach. We identified a central point in each zone and used a spin-the-bottle approach to select the starting point for enumeration.24 Individual residences were then selected at a predefined spacing interval, which was based on the estimated population of each catchment area. The residences were sampled in a clockwise direction until the whole zone had been canvassed. Team members then went to the next zone on the list and started the process anew. In the first survey, sampling continued until 387 eligible households, as defined below, had been surveyed in each community. If the target number was reached midway through a zone, sampling was completed for that zone. The second survey was conducted in the same zones, used the same predefined residence spacing intervals and took place between March and December 2011, at least two years after the PMTCT programme had been introduced at each health-care facility. However, since the starting point for canvassing households in each zone was randomly selected, the two surveys did not necessarily include the same households.

At each household, trained enumerators identified the head of the household and administered a screening questionnaire. If a household member was reported to have given birth in the last two years, written consent was requested for the use of an additional questionnaire with 165-questions to collect more data on the demographic and socioeconomic characteristics of the household and on the medical history of the child’s mother. Separate written consent was also requested for the collection of blood specimens from eligible mothers and children. Maternal samples were tested for HIV antibodies and specimens from children who were exposed to HIV (i.e. because their mothers were seropositive) were tested for HIV deoxyribonucleic acid using polymerase chain reaction. In households in which a newborn child had died within the past two years, survey staff administered a comprehensive verbal autopsy questionnaire.25 If members of a selected household were not available at the time of the initial visit, enumerators returned up to three times and scheduled appointments to meet with either the head of the household or the child’s mother or both.

The primary outcome measure was the proportion of HIV-exposed children that were alive and HIV-uninfected at 24 months of age (i.e. 24-month HIV-free survival). We hypothesized that the proportion would increase between the two surveys from 50% to 75%.26 To achieve a power of 80% with an α of 0.05 using the χ2 test, we needed to enrol 58 HIV-exposed children born within the last two years from each of the four study sites. Assuming an estimated prevalence of HIV infection of 15%, we calculated that we needed to include a minimum of 387 households with a child under the age of two years in each community.

Statistical methods

We compared participants in the two surveys. Differences in categorical variables were assessed using Pearson’s χ2 test after we confirmed that the test’s assumptions were valid in each case. Differences in continuous variables were assessed using nonparametric Wilcoxon rank-sum tests. The overall HIV-free survival rate in HIV-exposed children was derived for each survey using a parametric survival model with a Weibull distribution. A Weibull model was considered appropriate following graphical exploration of the baseline hazard. We also examined differences in HIV-free survival between the two surveys in each individual community using the same approach. Further, we identified factors associated with HIV infection or death by calculating adjusted hazard ratio (aHR) using Weibull regression with interval censoring and adjustment for clustering. The variables selected a priori for inclusion in our multivariable models were the PMTCT therapy, maternal age, parity, educational level and marital status, institutional antenatal care, institutional delivery and infant breastfeeding. In addition, a binary variable for whether the data were collected in the first or second household survey (i.e. before or after programme implementation) was included to take into account unmeasured time trends between the surveys. Because extensive information was collected in the household questionnaires, the multivariable models included socioeconomic and children’s health factors that were associated with HIV-free survival at a statistical significance level of P < 0.10. A factor was excluded if it did not appear to be independent of other factors included in the proposed model. We controlled for the survey site using the strata option in Stata version 12.1 (StataCorp. LP, College Station, United States of America). Further, to correct for possible correlations within sites, we employed a sandwich estimator to obtain robust standard errors using the vce(robust) option. In addition, because we were interested in the detailed effect of the pilot PMTCT programme on programme outcomes, we used Pearson’s χ2 test to compare the utilization of specific PMTCT services between respondents who took part in the first and second surveys. For all analyses, we considered P < 0.05 to be statistically significant.

The study was approved by the Biomedical Research Ethics Committee of the University of Zambia and the institutional review boards of the University of North Carolina at Chapel Hill and of the University of Alabama at Birmingham in the United States.

Results

In total, 31 zones were visited in each of the two surveys: seven in Chipapa, six in Kafue Estates, nine in Kafue Missions and nine in Mount Makulu. In the first survey, 3636 households were approached, 1947 (53.5%) of which were found to be eligible (Fig. 1). Information about maternal HIV status was available for 1778 (97.2%) of the 1830 children aged under two years in these households; 335 (18.8%) were found to be HIV-exposed. In the second survey, 5801 households were approached, 2441 (42.1%) of which were found to be eligible. Information about maternal HIV status was available for 2386 of 2444 (97.6%) children aged under two years and 390 (16.3%) were found to be HIV-exposed. The geographical distribution of eligible households sampled in the second survey is shown in Fig. 2 – comparable data were not collected in the first survey.

Fig. 1. Community surveys before and during the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 1. Community surveys before and during the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
a The first survey was conducted between November 2008 and May 2009, before implementation of the PMTCT pilot programme.
b The second survey was conducted between March and December 2011, at least two years after the PMTCT pilot programme had been introduced.
Fig. 2. Geographical distribution of eligible households in the second community surveya during the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 2. Geographical distribution of eligible households in the second community survey<sup>a</sup> during the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
a Comparable data were not collected in the first survey.
Source: Map produced using ArcGIS® (Esri, Redlands, United States of America) and shapefile of roadmap.27

There were significant differences in household and maternal characteristics between the two surveys (Table 2). Socioeconomic conditions were better in the second survey than the first: households in the second survey were significantly more likely to report having a water supply, a finished floor, electricity, a refrigerator, a television and a mobile phone (P < 0.0001 for all). The prevalence of HIV infection among mothers appeared to decline between the surveys, from 18.8% to 16.3% (P < 0.071), and the proportion of women who gave birth at a health-care facility increased from 58.0% to 67.8% (P < 0.0001). The children’s’ characteristics appeared similar in the two surveys. In particular, there was no significant difference in their median age: 10.9 months in the first survey versus 10.8 months in the second (P = 0.081, Table 2).

The cumulative estimated 24-month survival in HIV-exposed children was significantly higher in the second survey than the first: 0.97 (95% confidence interval, CI: 0.95–0.99) versus 0.87 (95% CI: 0.84–0.95), respectively (Fig. 3). The difference in the estimated 24-month HIV-free survival was even more pronounced: 0.89 (95% CI: 0.83–0.94) in the second survey versus 0.66 (95% CI: 0.63–0.76) in the first (Fig. 4). These trends were consistent across all four communities. Multivariate analysis showed that the risk of HIV infection or death was significantly lower in the children of mothers who started a combination antiretroviral regimen during pregnancy than in those whose mothers had no antiretroviral prophylaxis (aHR: 0.33, 95% CI: 0.15–0.73). Moreover, even after adjustment for potential demographic and socioeconomic confounders, children in the second survey were less likely to acquire an HIV infection or die than those in the first (aHR: 0.52, 95% CI: 0.34–0.80; Table 3).

Fig. 3. Estimated survival of children born to HIV-infected mothers before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 3. Estimated survival of children born to HIV-infected mothers before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
CI: confidence interval.
Fig. 4. Estimated HIV-free survival of children born to HIV-infected mothers before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 4. Estimated HIV-free survival of children born to HIV-infected mothers before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
CI: confidence interval.

To assess the contribution of the pilot PMTCT programme to the observed improvement in the HIV-free survival rate in children, we examined differences in the utilization of specific PMTCT services between the surveys. Increases were observed between the first and second surveys in four key indicators in HIV-infected mothers (i.e. those who tested positive at the time of the survey): (i) the proportion who knew their HIV status (from 74.4% to 91.9%, P < 0.0001; Fig. 5, available at: http://www.who.int/bulletin/volumes/92/8/13-129833); (ii) the proportion who were tested for HIV during their last pregnancy (from 76.4% to 88.1%, P < 0.0001; Fig. 6, available at: http://www.who.int/bulletin/volumes/92/8/13-129833); (iii) the proportion who reported using any antiretroviral prophylaxis during their last pregnancy (from 42.5% to 67.5%, P < 0.0001; Fig. 7, available at: http://www.who.int/bulletin/volumes/92/8/13-129833); and (iv) the proportion who reported using combination antiretroviral regimens during their last pregnancy (from 13.3% to 45.3%, P < 0.0001; Fig. 8).

Fig. 5. Proportiona of HIV-infected mothers who knew their HIV status, by health-care facility, before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 5. Proportion<sup>a</sup> of HIV-infected mothers who knew their HIV status, by health-care facility, before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 6. Proportiona of HIV-infected mothers tested for HIV during their last pregnancy, by health-care facility, before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 6. Proportion<sup>a</sup> of HIV-infected mothers tested for HIV during their last pregnancy, by health-care facility, before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 7. Proportiona of HIV-infected mothers who reported using any antiretroviral prophylaxis during their last pregnancy, by health-care facility, before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 7. Proportion<sup>a</sup> of HIV-infected mothers who reported using any antiretroviral prophylaxis during their last pregnancy, by health-care facility, before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
ART, antiretroviral therapy.
Fig. 8. Proportiona of HIV-infected mothers who reported using a combination antiretroviral regimen during their last pregnancy, by health-care facility, before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011
Fig. 8. Proportion<sup>a</sup> of HIV-infected mothers who reported using a combination antiretroviral regimen during their last pregnancy, by health-care facility, before and after the pilot programme to prevent mother-to-child transmission (PMTCT) of the human immunodeficiency virus (HIV) in the Kafue district of Zambia, 2009–2011

Discussion

Our study used before and after household surveys to assess the changes, at the population level, associated with a pilot PMTCT programme similar to WHO’s Option B. We observed a dramatic increase in the 24-month HIV-free survival rate among children born to HIV-infected mothers in our four targeted communities. However, because our study was not randomized and did not include a control group, we were not able to attribute this improvement to the provision of combination antiretroviral regimens to HIV-infected pregnant women in our pilot programme. Nevertheless, our findings indicate that increased investment in PMTCT can have a positive effect on early childhood health outcomes at the population level even though it was difficult to quantify the influence of individual components of the programme.

Our study had several other limitations in addition to the absence of randomization and the lack of a control group. First, the significant differences observed between our two survey populations, particularly in household indicators of socioeconomic status, were unexpected and could have contributed to a general improvement in health. Although we attempted to adjust for these factors in our multivariate analysis, there remained a risk of residual confounding. Moreover, other studies carried out in these communities between 2008 and 2011 may also have contributed to improved health outcomes.2830 Second, the number of HIV infections and deaths among HIV-exposed children was relatively small. However, the use of statistical models to estimate the 24-month HIV-free survival rate in children increased the precision of our estimates. Finally, in this analysis we did not explore reasons for the lower than expected uptake of maternal ART during pregnancy in the four communities. We plan to carry out a secondary analysis using the survey data to address this issue. However, given the likely heterogeneity of participants’ characteristics between the study communities, more extensive qualitative research may be needed.

Although we observed a more than threefold increase in the utilization of combination antiretroviral regimens between the first and second surveys, the rates were modest, at 13.3% and 45.3%, respectively. Since fewer than half of HIV-infected pregnant women received this intervention, it is unlikely that the pilot PMTCT programme alone could have produced such a dramatic increase in HIV-free survival in children. Better survival could, however, be explained by increased coverage of PMTCT and HIV treatment services in our target communities. It is possible that investment at our pilot sites may have resulted in both a greater supply of PMTCT services and an increase in demand. Indeed, we observed that the proportion of HIV-infected women who knew their HIV status during pregnancy increased between the two surveys, as did the proportion who reported HIV testing during their last pregnancy and the proportion that started any form of antiretroviral prophylaxis. It is certainly plausible that community outreach efforts and additional human resources associated with the pilot programme benefited all HIV-infected pregnant women, whether or not they eventually started combination antiretroviral regimens. This broader influence of PMTCT services should be considered when national programmes seek to scale up the implementation of Option B or Option B+.31 Each step of the PMTCT cascade must be addressed at the health systems level if we are to maximize the effect of interventions implemented during pregnancy and breastfeeding.32

This study assessed the effectiveness of the pilot PMTCT programme using community-based household surveys, which is one of five approaches endorsed by WHO.33 However, we encountered several unanticipated challenges. The proportion of eligible households that refused to participate in the first survey was much higher than in the second: 11.3% versus 1.6%, respectively (Fig. 1). The number of refusals declined at each of the four study sites; the greatest difference was noted in the Kafue Missions catchment area, where the proportion decreased from 22.2% to 1.2%. There was anecdotal evidence that our first survey coincided with the circulation of negative rumours about clinical research, which may explain the high refusal rate. It is important, therefore, that extensive, sustained outreach work be carried out during study implementation. We failed to collect information about which mothers participated in both surveys. However, the overlap is likely to be low because the latest demographic and health survey in Zambia indicated that only 15.2% of women with a recent pregnancy reported the delivery of another child within the previous 24 months.34 Moreover, since our surveys took place more than two years apart, no HIV-exposed child aged under 24 months could have been included in both surveys. In addition, we acknowledge that our estimates of HIV-free survival have not been validated against a gold standard metric, as could be achieved by longitudinal follow-up of HIV-exposed infants over time. Currently, this is being remedied in a study funded by the United States National Institutes of Health (clinicaltrials.gov ID: NCT01951794). While the precision of our estimates for HIV-free survival requires confirmation, we believe our comparisons are valid because we used the same methods in both surveys.

In conclusion, this pilot PMTCT programme in rural Zambia was associated with an increase in the 24-month HIV-free survival in children born to HIV-infected mothers; however, we were unable to quantify the contribution of specific programme features to health outcomes. Increased investment at the pilot sites probably contributed to the observed improvements in health by increasing both the supply of and demand for PMTCT services. Countries planning to incorporate WHO’s Option B or Option B+ strategy for PMTCT into their national PMTCT policies should bear in mind the importance of health systems capacity to support the effective incorporation of more efficacious regimens.


Funding:

The study was funded by a Clinical Scientist Development Award from the Doris Duke Charitable Foundation (2007061). Additional trainee support was provided by the National Institutes of Health through the International Clinical Research Scholars Program at Vanderbilt University (R24 TW007988).

Competing interests:

None declared.

References

Share